The ‘Referendum Bubble’: what can we learn from EU campaign polling? – EU Referendum Analysis 2016

The ‘Referendum Bubble’: what can we learn from EU campaign polling?

With the EU Referendum coming just thirteen months after a General Election in which the predictions were proven so dramatically wrong, pollsters were more cautious about publicising polls over the course of the Referendum campaign. The final figures from YouGov (52-48), Populus (55-45) and Com Res (48-42) may have disagreed on the extent of the victory, but in a binary choice they all had Remain comfortably across the line. With a recent history of methodological miscalculations and disagreement between the polls on the exact lead for Remain perhaps we should have taken these findings with a pinch more salt.

… the proportion of voters who felt that Remain would win never dropped below 54% while the number who felt that Leave would win was never higher than 26%.  This may possibly explain why Remain voters were less likely to go to polling stations… (and) why so many Leave voters expressed dismay at the result– they had more incentive to turn out, but they never actually expected to win.

When the result was announced, YouGov stated that its miscalculation was due to a much higher turnout in those areas coming out in favour of a Leave vote. And in the days following the vote, the media began to focus on how David Cameron and his team had been isolated in a Referendum bubble, supremely confident in their own success and oblivious to the way that the campaign was really going. There were reports of premature celebrations among the Prime Minister and his Downing Street advisers that were quite unlike the 2015 General Election when David Cameron had reportedly written and even practised a resignation speech. Cameron hadn’t felt the need to do the same on Referendum night.

Predicting Referendum results is even more difficult than predicting General Election results, and estimating turnout is particularly complex. But if we draw on political science, two things should have been clear:

1. We should have expected a higher turnout among Leave voters: those in favour of change tend to be more likely to vote than those in favour of maintaining the status quo. Polling in early 2016 clearly demonstrated that those in favour of Brexit were more likely to say they would definitely vote. So the fact that an even larger number of those who supported the Leave campaign actually went to the polling station should not come as too much of a surprise.

2. There was less incentive for Remain voters to go out and vote: we know that voters see the process of voting as collective action, but we also know from calculus models of voting that there is more incentive for people to turnout if they think that their own vote may actually make a difference to the outcome. Most of the polls through the latter parts of the campaign predicted that the outcome would be close. But polling gave a sense that the Remain campaign would be victorious. Polling commissioned by The Telegraph, for instance, showed the proportion of voters who felt that Remain would win never dropped below 54% while the number who felt that Leave would win was never higher than 26%. This may possibly explain why Remain voters were less likely to go to polling stations. They may have perceived that their vote would make less of an impact on the overall result. It also explains why so many Leave voters expressed dismay at the result– they had more incentive to turn out, but they never actually expected to win.

It seems then, that it wasn’t just Downing Street that was in a Referendum bubble. The pollsters, the media and the majority of the public were too. And on Friday 24th June that bubble finally burst. But looking at polling on British attitudes towards the European Union is useful for other reasons. In fact, if we consider polls from the last two Parliaments we can see two things which help to put the Referendum result into context:

1. The vote for Leave was not as strong as it has been in the past: YouGov polling specifically on EU membership since 2010 shows that the gap between those in favour of leaving and those who would prefer to remain has been as large as 23 points. This was in September 2012, in the middle of the Greek bailout crisis. Since 2013 the gap between the two sides narrowed considerably, with support for leave and remain regularly overlapping before the Referendum was formally announced.

2. Although there has always been a lingering Euroscepticism, attitudes to the EU have historically seen large swings. We can see this from polling data discussed above and we can also see it from the behaviour of our parliamentarians. It changed in line with milestones in the development and integration of the European Union, the accession of new member states, and changes in economic or political situations across Europe. It is likely then, that had the Remain side seen victory this week, attitudes towards the EU would have slowly picked up again.

Perhaps we should have foreseen this polling mishap. As last year’s General Election result showed us, even the best polling can get the result – not just its magnitude – wrong. Those who rely on polling are always taking a gamble if they don’t understand some of its biggest limitations.